Pinned Loading

1. llama.cpp-1-bit-turbo llama.cpp-1-bit-turbo Public

   Forked from ggml-org/llama.cpp

   HIP/ROCm fork optimized for AMD RDNA2 (gfx1030) with PrismML Q1_0_G128 1-bit quant support, RotorQuant, TurboQuant, EAGLE3 and P-EAGLE speculative decoding, and full Wave32 kernel optimizations.

   C++ 5

2. sglang-1-bit-turbo sglang-1-bit-turbo Public

   Forked from sgl-project/sglang

   SGLang 1-Bit Turbo — AMD ROCm (gfx1030) inference fork with RotorQuant/TurboQuant KV compression, PHANTOM-X zero-copy draft speculation, EAGLE3 speculative decoding, 12 RDNA2 crash fixes, and Prism…

   Python 4

3. vllm-1-bit-turbo vllm-1-bit-turbo Public

   Forked from mitkox/vllm-turboquant

   vLLM 0.18.1rc1 fork optimized for HIP/ROCm with support added for PrismML Bonsai Q1_0 and Q1_0_G128 1-bit GPU inference, TurboQuant TQ3_0 KV cache, and AMD gfx1030/RDNA2 architecture.

   Python

4. gfxGRAPH gfxGRAPH Public

   CUDA Graph → HIP Graph translation layer for AMD gfx1030 (RDNA2). Bridges all 4 CUDA Graph parity gaps on ROCm.

   Python

5. SpecForge SpecForge Public

   Forked from sgl-project/SpecForge

   Train speculative decoding models effortlessly and port them smoothly to SGLang serving.

   Python

6. ATLAS ATLAS Public

   Forked from itigges22/ATLAS

   Adaptive Test-time Learning and Autonomous Specialization

   Python